Centralize. Control. Consume: The AI Playbook

Technological revolutions rarely arrive as ruptures. They come as repetitions—patterns that echo past cycles of promise, consolidation, and control. The rise of artificial intelligence is no exception. Beneath the rhetoric of “democratization” and “empowerment” lies a familiar playbook: Centralize. Control. Consume. It is a strategy that has shaped every major technological epoch from the telegraph to the internet, and it is now being rewritten for the age of AI.

Each wave of digital innovation begins with openness and ends in consolidation. The early internet promised decentralization—a “network of peers,” in the words of its architects—but soon gave way to powerful intermediaries. Google’s search algorithms evolved from neutral tools of discovery into gatekeepers of attention. Facebook transformed from a social connector into an engine for behavioral data extraction. Amazon’s marketplace, once a platform for small sellers, became a mechanism for self-preference and exclusion.

These firms perfected the modern cycle of dominance: scale quickly, capture infrastructure, then extract value. What began as democratization ended as enclosure. The same arc now defines AI. Leading firms—OpenAI, Google DeepMind, Anthropic, Microsoft—deploy the language of accessibility while constructing proprietary models, gated APIs, and closed ecosystems. The public narrative is openness; the operational reality is control.

The new AI giants are not inventing a new model; they are following the old one. Like their predecessors in search or social media, they exploit the properties of the networked world—zero marginal cost, global reach, and data accumulation—to scale at unprecedented speed. Once scale is achieved, the logic of centralization takes over: restrict access, limit interoperability, and subsume competitors.

History offers precedents. AT&T’s telephone monopoly in the early 20th century justified its dominance as “universal service,” even as it stifled competition. IBM’s mainframe empire in the 1960s promised efficiency but imposed dependency. Facebook, too, acquired nascent competitors such as Instagram and WhatsApp not solely to expand offerings but to neutralize threats to its attention monopoly. Today’s AI rhetoric--democratization through APIs—follows the same logic. Beneath the veneer of access lies the consolidation of computational capital.

Yet two features distinguish this moment from earlier ones.

First, the geopolitical dimension. The United States no longer holds a monopoly on innovation. China now rivals Silicon Valley in AI research, cloud infrastructure, and surveillance applications. This competition intensifies the drive for scale, secrecy, and speed. AI is not merely a technological race; it is a struggle for global influence, where openness becomes a liability.

Second, the assertion of agency from the Global South. During the rise of Web 2.0, much of the world served as a market for Western platforms. That is no longer the case. Researchers and governments across Africa, Latin America, and South Asia now insist on participation in shaping datasets, linguistic models, and ethical frameworks. This shift challenges the one-directional flow of technology and knowledge that characterized the first internet era. It also introduces a critical moral dimension: technological sovereignty is increasingly framed not as protectionism, but as empowerment.

The deeper concern is cultural, not merely economic. The early web empowered users to build: to code, share, remix, and publish. Over time, that agency was eroded. Algorithms began curating what people saw, thought, and bought. Users became data points in systems optimized for engagement rather than understanding.

AI intensifies this passivity. It does not merely filter experience—it produces it. When systems draft our emails, write our code, and generate our art, the boundary between tool and author blurs. As dependence deepens, human initiative contracts. This is not an inevitable outcome of AI, but the result of incentive structures that reward control over creativity.

Moreover, the centralization of AI has implications for knowledge itself. Historically, innovation thrived where experimentation was distributed and localized. Bell Labs and Xerox PARC, for example, succeeded because researchers had autonomy and access to open experimentation spaces. Today, centralized AI platforms risk replicating the opposite model: research agendas dictated by the priorities of a few corporate decision-makers, shaping what is studied, published, and deployed globally.

The business model behind AI echoes that of earlier monopolies: subsidize access, capture users and data, and then restrict alternatives. Standard Oil did it with pipelines; Microsoft with operating systems; Google with information. The invisible bars today are data lock-ins, model dependencies, and ecosystem exclusivity. Once entrenched, such systems are difficult to escape because they are designed not just to serve but to enclose.

Reclaiming Agency

AI is not inherently monopolistic. Its underlying technologies—open-source models, federated learning, community datasets—could enable distributed power and creativity. The question is whether these alternatives can survive the gravitational pull of capital and scale.
History suggests they can, but only with deliberate intervention. The rise of open-source software in the 1990s showed that collaborative models can compete with proprietary systems. Similarly, decentralized AI initiatives—from collaborative model hubs to community-governed datasets—point to a potential future where users reclaim agency over their tools and knowledge.

The repetition of the “centralize, control, consume” cycle is not inevitable. History demonstrates that incentives, regulations, and technical architecture shape whether technology empowers or encloses. To counter monopolistic concentration in AI, three complementary strategies are critical: governance, interoperability, and participatory design.

1. Governance and Oversight. Governments and independent regulatory bodies can establish frameworks that prevent monopolistic consolidation while encouraging innovation. Just as antitrust actions against AT&T and Microsoft in the late 20th century created space for competition, policymakers today could enforce transparency in AI model ownership, data usage, and commercial exclusivity. International coordination is also essential: AI is global, and unilateral regulation risks ceding influence to actors who enforce weaker standards, whether corporations or states; in practice, this may require a patchwork approach—alliances of like-minded countries, interoperable regional frameworks, and shared technical standards that raise the floor globally.

2. Interoperability and Open Standards. Centralization thrives on lock-in. Mandating open APIs, standardized model formats, and cross-platform compatibility can reduce dependency on a single provider. Historical precedents include the early internet protocols (TCP/IP, HTML) that enabled broad participation and competition. Federated AI models, which allow local training while sharing knowledge across nodes, offer a technical analog: users retain control while benefiting from collective intelligence.

3. Participatory Design and Local Agency. The Global South’s insistence on inclusive AI demonstrates the importance of context-sensitive systems. Community-governed datasets, multilingual models, and local ethical oversight ensure AI tools reflect diverse values, rather than reproducing centralized norms. Early open-source software projects exemplify how distributed collaboration can generate robust, widely adopted technology while avoiding enclosure.

These interventions are mutually reinforcing. Governance provides the guardrails, interoperability reduces lock-in, and participatory design empowers users as creators rather than passive consumers. Combined, they chart a path for AI to realise the early internet’s promise: a tool for collective creativity, knowledge, and agency.

The Cage and the Playground
Every technological era creates both playgrounds and cages. The early internet was a playground because users were active co-creators—they could build websites, share code, form communities, and shape the norms and culture of the web. Platforms like Geocities, early forums, and open-source projects empowered ordinary people to participate in ways previously unimaginable. Knowledge, creativity, and social connection were decentralised, flowing in multiple directions rather than being filtered through a few gatekeepers.

But this playground became a cage as control migrated upward. Commercial incentives, centralized platforms, and algorithmic curation gradually siphoned agency away from users. Google, Facebook, and Amazon did not start as monopolies, but their architectures and business models concentrated power, reducing users to passive participants who scrolled, clicked, and consumed content dictated by attention-maximizing systems. What had been a generative space of co-creation became a curated ecosystem of consumption.

The AI era faces the same crossroads, but with higher stakes. Unlike the early web, AI systems can anticipate our decisions, generate content, and even influence thought and behavior. They promise convenience and efficiency, but this very promise risks displacing human judgment, experimentation, and initiative. The question, then, is not what AI can do, but who it serves. Will it amplify human creativity and decision-making, or will it cement centralized authority under the guise of assistance?
​
If left to the same incentives that governed Big Tech’s rise, the pattern will repeat: scale fast, centralize ruthlessly, and control relentlessly. But history also offers a roadmap for a different outcome. By insisting on openness, transparency, interoperable standards, and mechanisms that return control to users, AI can become the first technology that genuinely breaks the cycle rather than repeats it. The choice is not technical—it is cultural, political, and ethical: whether we use AI to expand human agency or allow it to quietly diminish the space in which we think, create, and participate.